/**
* 提取网页中的删除操作
*
* @param nodeList
* @param dList
* @return
*/
public List<DeleteEvent> processDelete(NodeList nodeList, List<DeleteEvent> dList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().contains("discussion-item-head_ref_deleted")) {
DeleteEvent d = new DeleteEvent();
// TODO 解析comment工作
Node deleteNode = DownloadUtil.getSomeChild(node, "span title=\"");
d.setRef(deleteNode.getText().split("\"")[1]);
System.out.println(deleteNode.getText().split("\"")[1]);
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"author\"");
d.setActor(actorNode.toPlainTextString());
System.out.println(actorNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
d.setDeleteAt(time);
}
dList.add(d);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processDelete(childList, dList);
}
}
}
return dList;
}
/**
* 提取网页中所有的IssueComment元素
*
* @param source
*/
private List<IssueCommentEvent> processComment(
NodeList nodeList, List<IssueCommentEvent> icList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().matches("div id=\"issuecomment-.*\".*+")) {
IssueCommentEvent i = new IssueCommentEvent();
// TODO 解析comment工作
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"author\"");
i.setActor(actorNode.toPlainTextString());
Node contentNode = DownloadUtil.getSomeChild(node, "div class=\"comment-body");
i.setCommentBody(contentNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
i.setCreatedAt(time);
System.out.println(time);
}
icList.add(i);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processComment(childList, icList);
}
}
}
return icList;
}
/**
* 处理开启pullrequest的需求
*
* @param nodeList
* @param pList
* @return
*/
public List<PullRequestEvent> processOpenPull(NodeList nodeList, List<PullRequestEvent> pList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().contains("div id=\"issue-")) {
PullRequestEvent pullRequestEvent = new PullRequestEvent();
pullRequestEvent.setAction("open");
Node commentNode = DownloadUtil.getSomeChild(node, "div class=\"comment-body");
pullRequestEvent.setBody(commentNode.toPlainTextString());
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"author");
pullRequestEvent.setActor(actorNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
pullRequestEvent.setCreatedAt(time);
}
pList.add(pullRequestEvent);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processOpenPull(childList, pList);
}
}
}
return pList;
}
public List<PullRequestReviewCommentEvent> processSubPullRequestReviewComment(
NodeList nodeList, List<PullRequestReviewCommentEvent> prList, String discussionId) {
SimpleNodeIterator sni2 = nodeList.elements();
while (sni2.hasMoreNodes()) {
Node node2 = sni2.nextNode();
if (node2.getText().contains("div id=\"discussion_r")) {
PullRequestReviewCommentEvent p = new PullRequestReviewCommentEvent();
// TODO 解析comment工作
p.setDiscussionId(discussionId);
Node actorNode = DownloadUtil.getSomeChild(node2, "class=\"author\"");
p.setActor(actorNode.toPlainTextString());
System.out.println(actorNode.toPlainTextString());
Node contentNode = DownloadUtil.getSomeChild(node2, "div class=\"comment-body");
p.setCommentBody(contentNode.toPlainTextString());
System.out.println(contentNode.toPlainTextString().trim());
Node timeNode = DownloadUtil.getSomeChild(node2, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
p.setCreatedAt(time);
}
prList.add(p);
} else {
// 得到该节点的子节点列表
NodeList childList = node2.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processSubPullRequestReviewComment(childList, prList, discussionId);
}
}
}
return prList;
}
public List<Newsitem> parseContent(String content) throws Exception {
List<Newsitem> newsitems = new ArrayList<Newsitem>();
Tag newsDiv = this.extractTagByClassName(this.stripHtmlComments(content), "box_news");
NodeList nodes = this.extractTagsByClassName(newsDiv.toHtml(), "subItem");
for (int i = 0; i < nodes.size(); i++) {
NewsitemImpl newsitem = new NewsitemImpl();
Tag itemTable = (Tag) nodes.elementAt(i);
Tag titleTag = this.extractTagByClassName(itemTable.toHtml(), "subItemtitle");
newsitem.setTitle(titleTag.toPlainTextString());
Node descriptionSpan =
titleTag.getNextSibling().getNextSibling().getNextSibling().getNextSibling();
newsitem.setDescription(
descriptionSpan
.toPlainTextString()
.replaceAll("[^\\u0000-\\u00FF]", " ")
.replace(" Read More...", "")
.trim());
Tag linkTag = (Tag) extractLinks(itemTable.toHtml(), "/index.php.*").elementAt(0);
newsitem.setUrl(URL_PREFIX + linkTag.getAttribute("href"));
newsitems.add(newsitem);
}
return newsitems;
}
/**
* 处理Reference了当前pullrequest的操作
*
* @param source
*/
public List<PullRequestEvent> processReference(NodeList nodeList, List<PullRequestEvent> pList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().contains("div class=\"discussion-item discussion-item-ref\"")) {
PullRequestEvent pullRequestEvent = new PullRequestEvent();
pullRequestEvent.setAction("ref");
Node anotherAtifactNode = DownloadUtil.getSomeChild(node, "class=\"title-link\"");
pullRequestEvent.setBody(
anotherAtifactNode == null ? "" : anotherAtifactNode.toPlainTextString());
Pattern artifactPattern = Pattern.compile("[a-zA-Z]+/[a-zA-Z]+/[a-zA-Z]+/[a-z[0-9]]+");
Matcher artifactMatcher =
artifactPattern.matcher(anotherAtifactNode == null ? "" : anotherAtifactNode.getText());
if (artifactMatcher.find()) {
String anotherAtifact = artifactMatcher.group();
pullRequestEvent.setPullrequestBaseRef(anotherAtifact);
System.out.println(anotherAtifact);
}
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"author\"");
pullRequestEvent.setActor(actorNode == null ? "" : actorNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
pullRequestEvent.setCreatedAt(time);
}
pList.add(pullRequestEvent);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processReference(childList, pList);
}
}
}
return pList;
}
public void testCompositeTagWithOneTextChild() throws ParserException {
String html = "<Custom>" + "Hello" + "</Custom>";
createParser(html);
CustomTag customTag = parseCustomTag(1);
assertEquals("child count", 1, customTag.getChildCount());
assertFalse("custom tag should not be xml end tag", customTag.isEmptyXmlTag());
assertEquals("starting loc", 0, customTag.getStartPosition());
assertEquals("ending loc", 8, customTag.getEndPosition());
assertEquals("starting line position", 0, customTag.getStartingLineNumber());
assertEquals("ending line position", 0, customTag.getEndingLineNumber());
Node child = customTag.childAt(0);
assertType("child", Text.class, child);
assertStringEquals("child text", "Hello", child.toPlainTextString());
}
/**
* 处理取消指派某人操作
*
* <p>跟之前一样，取消指派的是后面的家伙
*
* @param nodeList
* @param pList
* @return
*/
private List<PullRequestEvent> processUnassigned(
NodeList nodeList, List<PullRequestEvent> pList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().contains("class=\"discussion-item discussion-item-unassigned\"")) {
PullRequestEvent pEvent = new PullRequestEvent();
pEvent.setAction("assigned");
Node assignedNode = DownloadUtil.getSomeChild(node, "class=\"author\"");
pEvent.setPullrequestAssgnee(assignedNode.toPlainTextString());
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"discussion-item-entity\"");
if (actorNode != null) {
pEvent.setActor(actorNode.toPlainTextString());
} else {
pEvent.setActor(assignedNode.toPlainTextString());
}
System.out.println(actorNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
pEvent.setCreatedAt(time);
}
pList.add(pEvent);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processUnassigned(childList, pList);
}
}
}
return pList;
}
private void processNodeList(NodeList list, List<String> valueList) {
// 迭代开始
SimpleNodeIterator iterator = list.elements();
while (iterator.hasMoreNodes()) {
Node node = iterator.nextNode();
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null == childList) {
// 得到值节点的值
String result = node.toPlainTextString().trim();
// 若包含关键字，则简单打印出来文本
// System.out.println(result);
if (result != null && !"".equals(result)) valueList.add(result);
} // end if
// 孩子节点不为空，继续迭代该孩子节点
else {
processNodeList(childList, valueList);
} // end else
} // end wile
}
/**
* 处理labeled操作
*
* @param source
*/
public List<PullRequestEvent> processLabled(NodeList nodeList, List<PullRequestEvent> pList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().contains("class=\"discussion-item discussion-item-labeled\"")) {
PullRequestEvent pullRequestEvent = new PullRequestEvent();
pullRequestEvent.setAction("labeled");
List<Node> lableList = new ArrayList<Node>();
lableList = DownloadUtil.getLableList(node, "style=\"color:", lableList);
String lables = "";
for (int i = 0; i < lableList.size(); i++) {
lables += lableList.get(i).toPlainTextString();
if (i != lableList.size() - 1) {
lables += ",";
}
}
System.out.println(lables);
pullRequestEvent.setPullrequestBaseLabels(lables);
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"author\"");
pullRequestEvent.setActor(actorNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
pullRequestEvent.setCreatedAt(time);
}
pList.add(pullRequestEvent);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processLabled(childList, pList);
}
}
}
return pList;
}
/**
* 处理移除里程碑动作
*
* @param nodeList
* @param pList
* @return
*/
public List<PullRequestEvent> processRemoveMileStone(
NodeList nodeList, List<PullRequestEvent> pList) {
SimpleNodeIterator sni = nodeList.elements();
while (sni.hasMoreNodes()) {
Node node = sni.nextNode();
if (node.getText().contains("div class=\"discussion-item discussion-item-demilestoned\"")) {
PullRequestEvent p = new PullRequestEvent();
p.setAction("removeMilestone");
Node milestoneNode = DownloadUtil.getSomeChild(node, "class=\"discussion-item-entity\"");
Pattern milestonePattern = Pattern.compile("[a-zA-Z]+/[a-zA-Z]+/[a-zA-Z]+/.*+");
Matcher milestoneMatcher = milestonePattern.matcher(milestoneNode.getText());
if (milestoneMatcher.find()) {
String milestone = milestoneMatcher.group().split("\"")[0];
p.setBody(milestone);
}
Node actorNode = DownloadUtil.getSomeChild(node, "class=\"author\"");
p.setActor(actorNode.toPlainTextString());
Node timeNode = DownloadUtil.getSomeChild(node, "datetime");
Pattern pattern = Pattern.compile("datetime=\".*\"");
Matcher matcher = pattern.matcher(timeNode.getText());
if (matcher.find()) {
String time = matcher.group().split("\"")[1];
p.setCreatedAt(time);
}
pList.add(p);
} else {
// 得到该节点的子节点列表
NodeList childList = node.getChildren();
// 孩子节点为空，说明是值节点
if (null != childList) { // 如果孩子结点不为空则递归调用
processRemoveMileStone(childList, pList);
}
}
}
return pList;
}
/** 从课表处，分课表 */
public List<Courses> parseCourses(String html) {
Parser parser = new Parser();
try {
parser.setInputHTML(html);
parser.setEncoding("utf-8");
} catch (ParserException e) {
e.printStackTrace();
}
NodeFilter filter = new NodeClassFilter(TableTag.class);
NodeList nodeList = null;
try {
nodeList = parser.extractAllNodesThatMatch(filter);
} catch (ParserException e) {
e.printStackTrace();
}
List<Courses> list = new ArrayList<Courses>();
String schoolyear = "";
String semester = "";
for (int i = 0; i < nodeList.size(); i++) {
if (nodeList.elementAt(i) instanceof TableTag) {
TableTag tag = (TableTag) nodeList.elementAt(i);
TableRow[] rows = tag.getRows();
for (int j = 0; j < rows.length; j++) {
TableRow row = (TableRow) rows[j];
TableColumn[] columns = row.getColumns();
Courses courses = null;
boolean isCourse = false;
for (int k = 0; k < columns.length; k++) {
Node columnNode = columns[k];
String info = columnNode.toPlainTextString().trim();
String temp = "学年学期：";
int start = info.indexOf(temp);
int len = "2012-2013".length();
if (start != -1) {
start = start + temp.length();
schoolyear = info.substring(start, start + len);
// semester = info.substring(start+len+2);
// 网络正常时候测试学期改为数字
semester = info.substring(start + len + 3, start + len + 4);
if ("一".equals(semester)) {
semester = "1";
} else if ("二".equals(semester)) {
semester = "2";
}
}
if (k == 1 && info.indexOf("[") != -1) {
courses = new Courses();
String courseCode = info.substring(1, 9);
String coursesname = info.substring(10);
courses.setCourseCode(courseCode);
courses.setCoursesname(coursesname);
isCourse = true;
}
if (k == 2 && isCourse) {
double credit = Double.parseDouble(info);
courses.setCredit(credit);
}
if (k == 3 && isCourse) {
courses.setType(info);
}
if (k == 4 && isCourse) {
courses.setLeanType(info);
}
if (k == 5 && isCourse) {
courses.setCheckType(info);
}
if (k == 6 && isCourse) {
courses.setGetType(info);
}
if (k == 7 && isCourse) {
// double score=Double.parseDouble(info);
courses.setScore(info);
}
if (k == 8 && isCourse) {
courses.setRemark(info);
}
} // end for k
if (courses != null) {
courses.setSchoolYear(schoolyear);
courses.setSemester(semester);
list.add(courses);
}
} // end for j
}
}
return list;
}
public List<TimeTable> parseTimeTables(String html) {
Parser parser = new Parser();
try {
parser.setInputHTML(html);
parser.setEncoding("utf-8");
} catch (ParserException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
List<TimeTable> list = new ArrayList<TimeTable>();
NodeFilter filter = new NodeClassFilter(TableTag.class);
NodeList nodeList = null;
try {
nodeList = parser.extractAllNodesThatMatch(filter);
} catch (ParserException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
for (int i = 0; i < nodeList.size(); i++) {
if (nodeList.elementAt(i) instanceof TableTag) {
TableTag tag = (TableTag) nodeList.elementAt(i);
if (tag.getText().indexOf("[课程号]") == -1) {
continue;
}
TableRow[] rows = tag.getRows();
for (int j = 1; j < rows.length; j++) {
TableRow row = (TableRow) rows[j];
TableColumn[] columns = row.getColumns();
boolean isCourse = false;
TimeTable timeTable = null;
for (int k = 0; k < columns.length; k++) {
Node columnNode = columns[k];
String info = columnNode.toPlainTextString().trim();
System.out.println(info + "===" + k);
switch (k) {
case 1:
int start = info.indexOf("[");
int end = info.indexOf("]");
timeTable = new TimeTable();
timeTable.setCourseCode(info.substring(start + 1, end));
timeTable.setCourseName(info.substring(end + 1));
break;
case 3:
timeTable.setCredit(Double.parseDouble(info));
break;
case 4:
timeTable.setType(info);
break;
case 5:
int t_start = info.indexOf("]");
timeTable.setTeacher(info.substring(t_start + 1));
break;
case 8:
List<TimeAndAdress> ta_list = praseStr(info);
for (TimeAndAdress ta : ta_list) {
timeTable.setAddress(ta.getAddress());
timeTable.setTime(ta.getTime());
timeTable.setCycle(ta.getCycle());
timeTable.setSingleDouble(ta.getSingleDouble());
timeTable.setWeek(ta.getWeek());
list.add(timeTable.clone());
}
break;
default:
break;
}
}
} // end for j
}
}
return list;
}
public List<TimeTable> parseTimeTables(String html) {
Parser parser = new Parser();
try {
parser.setInputHTML(html);
parser.setEncoding("utf-8");
} catch (ParserException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
List<TimeTable> list = new ArrayList<TimeTable>();
NodeFilter tagfilter = new NodeClassFilter(TableTag.class);
NodeFilter idFilter = new HasAttributeFilter("id", "reportArea");
NodeFilter filter = new AndFilter(tagfilter, idFilter);
NodeList nodeList = null;
try {
nodeList = parser.extractAllNodesThatMatch(filter);
} catch (ParserException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
for (int i = 0; i < nodeList.size(); i++) {
if (nodeList.elementAt(i) instanceof TableTag) {
TableTag tag = (TableTag) nodeList.elementAt(i);
TableRow[] rows = tag.getRows();
for (int j = 0; j < rows.length; j++) {
TableRow row = (TableRow) rows[j];
TableColumn[] columns = row.getColumns();
boolean isCourse = false;
TimeTable timeTable = null;
for (int k = 0; k < columns.length; k++) {
Node columnNode = columns[k];
String info = columnNode.toPlainTextString().trim();
// System.out.println(info+"=="+k);
if (k == 1 && info.indexOf("[") != -1) {
timeTable = new TimeTable();
String courseCode = info.substring(1, 9);
String coursesname = info.substring(10);
timeTable.setCourseName(coursesname);
timeTable.setCourseCode(courseCode);
isCourse = true;
}
if (k == 2 && isCourse) {
double credit = Double.parseDouble(info);
timeTable.setCredit(credit);
}
if (k == 3 && isCourse) {
timeTable.setType(info);
}
if (k == 4 && isCourse) {
timeTable.setTeacher(info);
}
if (k == 5 && isCourse) {
timeTable.setClassId(info);
}
if (k == 6 && isCourse) {
timeTable.setClassNum(info);
}
if (k == 11 && isCourse) {
List<TimeAndAdress> ta_list = praseStr(info);
for (TimeAndAdress ta : ta_list) {
timeTable.setAddress(ta.getAddress());
timeTable.setTime(ta.getTime());
timeTable.setCycle(ta.getCycle());
timeTable.setSingleDouble(ta.getSingleDouble());
timeTable.setWeek(ta.getWeek());
list.add(timeTable.clone());
}
}
} // end for k
} // end for j
}
}
return list;
}
public ContentModel listHtml(String param, String type) {
ContentModel model = new ContentModel();
StringBuffer html = new StringBuffer();
try {
NodeFilter filter = new TagNameFilter("body");
Parser parser = new Parser();
parser.setURL(SearchHelper.SEARCH_URL_BAIDU + param);
parser.setEncoding(parser.getEncoding());
NodeList list = parser.extractAllNodesThatMatch(filter);
String body = list.toHtml();
Parser content = new Parser();
content.setInputHTML(body);
content.setEncoding(parser.getEncoding());
NodeFilter content_filter = new TagNameFilter("table");
NodeList content_list = content.extractAllNodesThatMatch(content_filter);
for (int i = 0; i < content_list.size(); i++) {
String s = content_list.elementAt(i).toHtml();
if (s.indexOf("div") != -1) {
continue;
}
if (s.indexOf("相关搜索") != -1) {
html.append("<div id=\"rs\">" + s + "</div>");
continue;
}
html.append("<div class=\"content\">");
for (Node n : extractHtml(content_list.elementAt(i), type)) {
if (n instanceof LinkTag) {
if (n.toPlainTextString().equals("百度快照")) {
continue;
}
html.append("<h3 class=\"t\">" + n.toHtml() + "</h3>");
} else {
html.append(n.toHtml());
}
}
html.append("<br/></div><br>");
}
/** 获取分页数据 */
Parser page = new Parser();
page.setInputHTML(body);
page.setEncoding(parser.getEncoding());
NodeFilter page_filter = new TagNameFilter("p");
NodeList page_list = page.extractAllNodesThatMatch(page_filter);
for (int i = 0; i < page_list.size(); i++) {
String s = page_list.elementAt(i).toHtml();
if (s.indexOf("page") == -1) {
continue;
}
html.append("<p id=\"page\">" + page_list.elementAt(i).toHtml() + "</div>");
}
} catch (Exception e) {
e.printStackTrace();
}
model.setContent(html.toString());
return model;
}